Struct QuantizationParams 

pub struct QuantizationParams {
    pub scale: f32,
    pub zero_point: i32,
    pub dtype: DType,
    pub qmin: i32,
    pub qmax: i32,
}

Quantization parameters for symmetric and asymmetric quantization

Fields

scale: f32

Scale factor for quantization (real_value = scale * quantized_value + zero_point)

zero_point: i32

Zero point for asymmetric quantization (usually 0 for symmetric)

dtype: DType

Quantized data type (Int8 or Int4)

qmin: i32

Minimum quantized value (e.g., -128 for INT8)

qmax: i32

Maximum quantized value (e.g., 127 for INT8)

Implementations

impl QuantizationParams

pub fn symmetric_int8(scale: f32) -> Self

Create symmetric INT8 quantization parameters

pub fn asymmetric_int8(scale: f32, zero_point: i32) -> Self

Create asymmetric INT8 quantization parameters

pub fn symmetric_int4(scale: f32) -> Self

Create symmetric INT4 quantization parameters

pub fn asymmetric_int4(scale: f32, zero_point: i32) -> Self

Create asymmetric INT4 quantization parameters

pub fn from_tensor_stats( min_val: f32, max_val: f32, dtype: DType, ) -> Result<Self>

Calculate quantization parameters from tensor statistics

pub fn asymmetric_from_tensor_stats( min_val: f32, max_val: f32, dtype: DType, ) -> Result<Self>

Calculate asymmetric quantization parameters from tensor statistics

Trait Implementations

impl Clone for QuantizationParams

fn clone(&self) -> QuantizationParams

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for QuantizationParams

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.